Applications of AI in Aquaculture Research and Farming

In recent years, artificial intelligence (AI) has become increasingly popular in aquaculture research and production. Both major industry players and startups are developing plans based on AI technology.

AI-enhanced fish farming systems. (Image: 123rf).

Tracing the Roots of AI

What we know about AI is that it is a programming technology related to image recognition, language processing, behavior analysis, and decision-making—without human supervision. Oxford Languages defines AI as “the theory and development of computer systems that perform tasks requiring human intelligence, such as visual perception, speech recognition, decision-making, and language translation.”

Salmon images captured by iFarm’s AI. (Image: thefishsite).

To understand the development of AI in aquaculture, one must mention Fuzzy Logic. Looking back to 1965, Lotfi A. Zadeh at the University of California, Berkeley published his research in a paper titled “Fuzzy Sets.” He introduced a unique concept related to “the theory of membership levels.” This laid the foundation for a mindset recognized as fuzzy logic.

While traditional computing relies on binary values of 0 and 1, true or false, fuzzy logic delves into the nuances of human perception, existing somewhere between 0 and 1, between true and false. Over the years, fuzzy logic has been applied in computing for image processing and medical diagnosis, paving the way for the robust development of AI technology we see today.

AI minimizing labor in aquaculture. (Image: deeptrekker).

AI technology has been developed and widely applied across various fields, including aquaculture production. The two most impactful technologies are deep learning and convolutional neural networks. Traditional machine learning uses algorithms to function based on provided data while improving through user feedback. In contrast, deep learning takes it a step further by encapsulating entire algorithms, self-evaluating, and self-adjusting. Convolutional neural networks are specialized deep learning algorithms particularly useful in image recognition and interpretation.

At the same time, sensor technology has made remarkable advancements, such as connectivity through Cloud systems, 5G networks, and the Internet of Things. Consequently, AI has been integrated into aquaculture to enhance feeding efficiency, assess biomass, monitor growth, detect diseases early, control and monitor environments (especially in Recirculating Aquaculture Systems – RAS), and reduce labor costs. With sensor and processing technology, routine tasks no longer require as much human labor as before, leading to improved animal welfare.

Underwater camera Orbit FHD Fixed from ScaleAQ – one of the most sensitive cameras in the world. (Image: scaleaq).

The Era of “AI Integration”

Early disease detection in aquaculture based on fish behavior and appearance has become a promising research area for AI applications. The costs associated with components of perfect AI systems for fish farming remain relatively high. However, with decreasing prices and new approaches emerging, more producers, even with limited resources, can possess essential AI systems.

Recently, Darapaneni and colleagues introduced a system for early disease detection, assisting fish farmers in selecting suitable management methods for their ponds. This system operates through a process where underwater cameras or similar sensors capture images, upload them to the Cloud, and send them to the responsible party for processing. The data is then classified and analyzed using a trained AI model. Thanks to advanced connectivity, the turnaround time is just a few minutes, allowing AI to process not just one but multiple ponds in a single day.

Salmon harvesting system. (Image: fishfarmingexpert).

The application of AI in automatic feeding systems for fish and shrimp has garnered significant attention from researchers. Chen and colleagues at Tsinghua University (Beijing, China) utilized a predictive formula for biomass based on a support vector machine model combined with convolutional neural networks, using real-time water quality data to predict optimal feeding needs and amounts for shrimp in RAS systems. Results showed an error margin of only 3.7%, significantly lower than manual feeding.

Monitoring and biomass verification is another area where AI systems are applied. Gonçalves and colleagues from the Federal University of Mato Grosso do Sul, Brazil, described the usefulness of convolutional neural networks for counting juvenile fish. This AI method allows for counting fish even when they are obscured by others while predicting the movements of juvenile fish.

Umitron, based in Tokyo, Japan, is focusing all resources on integrating AI into aquaculture applications. Umitron’s system leverages real-time observations of swimming behaviors of livestock to determine feeding times and the necessary amounts of feed for each cage. This method significantly improves feed conversion ratios while reducing waste and logistics demands.

Global salmon producer Cermaq has also researched and developed an AI system over the past few years. The most notable model is iFarm, developed in partnership with technology partner BioSort, aiming to enhance the health and welfare of farmed fish in net cages. Initially, the company focused on fish interactions with the system, followed by testing to refine and optimize system functions and operations. Currently, the company is evaluating sensor technology, data collection and processing, as well as algorithms. Additionally, developing methods for sorting fish in net cages remains a top priority.

• The first woman in the world marries a 3D AI image
• With just one command, the new AI will create realistic videos like this
• Chinese scientist creates the world’s first “AI child”